Parameter estimation in the presence of non-Gaussian noise

A method for parameter estimation is derived that is insensitive to the noise distribution, and an example of its use for nonlinear systems is given. The method combines the sensitivity of the maximum-likelihood parameter estimator with the robustness of order statistics to reduce estimation uncertainty significantly, with only a slight increase in the variance. This algorithm shows improvements over conventional parameter estimates, in particular, in the case of small data sets.